Researcher Michael Mahoney

Researcher Biography

http://www.stat.berkeley.edu/~mmahoney/

Past Projects

Improving OOD Generalization Through Metric-informed Weight-space Augmentation and Architecture Search

Measuring Prediction Trustworthiness and Safety Through Neural Network Loss Landscape Analysis

Hessian Aware Neural Cleansing: Searching for Optimal Trade-offs between Adversarial Robustness, Accuracy, and Speed

Systematic Study of Neural Network Robustness Against Adversarial Attacks

Trust-Region Based Robustness of Neural Networks in the Face of Adversarial Attacks

Efficient Neural Networks through Systematic Quantization

Implementing and Evaluating Second-Order Optimization for Deep Learning on Diverse Architectures

Improving the Scaling of Deep Learning Networks by Characterizing and Exploiting Soft Convexity